Data-Driven Meteorological Forecasting For Divisional Cities In Bangladesh: An ANN- Based Climate Study

Abstract

Bangladesh is highly susceptible to the impacts of climate change. In Bangladesh, variations in temperature, rainfall, relative humidity (RH), wind speed, cloud coverage and sunshine duration can strongly influence agriculture, public health, and socio-economic conditions. Accurate city-level climate predictions are essential for informed planning. Yet most existing research focuses on national-level trends often overlooking local variability. This study aims to bridge that gap by forecasting climate parameters for eight divisional cities of Bangladesh for the period 2023–2042. Two machine learning techniques Artificial Neural Networks (ANN) and Recurrent Neural Networks (RNN) were applied to predict maximum temperature, rainfall and relative humidity with model performance assessed via MAE, MSE, RMSE, and R². ANN emerged as the most reliable model for these predictions.  The results reveal distinct spatial patterns. Cities including Khulna, Dhaka, Rajshahi, Rangpur, and Chittagong show an increase in maximum temperatures and reductions in rainfall and relative humidity. Conversely, Sylhet and Barishal exhibit declining temperatures while rainfall and humidity rise. On the other hand, Mymensingh demonstrates simultaneous increases in all three parameters. These findings underscore that climate change effects are heterogeneous across Bangladesh. Understanding these patterns is crucial for developing effective regional adaptation strategies. Areas experiencing rising temperatures and reduced rainfall may face heightened heat stress and water scarcity. Whereas wetter regions may encounter increased flooding and humidity related challenges. This research provides city specific data-driven insights that can assist policymakers, researchers and planners in designing targeted interventions to mitigate climate risks.

GANITJ. Bangladesh Math. Soc. 46.3 (2026) 030–046